To Mme or Not to Mme, That's the Question

Ensembles are becoming more and more critical in the forecast process. In particular for longer outlooks, the ensemble mean has more skill than a deterministic model. In general, ensembles provide critical information on the uncertainty in the forecasts. Recently an apparent paradox has been developing. On the one hand, Multi Model Ensembles (MMEs) tend to show more skill than single model ensembles. On the other hand, there is a drive toward unified modeling as a better business model for rapid development of the quality of operational models. In the age of evidence-based decision making, how do you deal with these conflicting directions? In order to make a true evidence-based decision with respect to MMEs, it is not sufficient to look at today's model behavior only. It is essential to model the accuracy of models along different development path. The present paper develops a simple model to predict model accuracy, which is then used to address where and when MMEs should be used.